Decart Python SDK

A Python SDK for Decart's models.

Installation

Using UV

uv add decart

Using pip

pip install decart

Documentation

For complete documentation, guides, and examples, visit: https://docs.platform.decart.ai/sdks/python

Quick Start

Image Editing (Process API)

import asyncio
import os
from decart import DecartClient, models

async def main():
    async with DecartClient(api_key=os.getenv("DECART_API_KEY")) as client:
        # Edit an image
        result = await client.process({
            "model": models.image("lucy-image-2"),
            "prompt": "Apply a painterly oil-on-canvas look while preserving the composition",
            "data": open("input.png", "rb"),
        })

        with open("output.png", "wb") as f:
            f.write(result)

asyncio.run(main())

Video Editing (Queue API)

For video editing jobs, use the queue API to submit jobs and poll for results:

async with DecartClient(api_key=os.getenv("DECART_API_KEY")) as client:
    # Submit and poll automatically
    result = await client.queue.submit_and_poll({
        "model": models.video("lucy-clip"),
        "prompt": "Restyle this footage with anime shading and vibrant neon highlights",
        "data": open("input.mp4", "rb"),
        "on_status_change": lambda job: print(f"Status: {job.status}"),
    })

    if result.status == "completed":
        with open("output.mp4", "wb") as f:
            f.write(result.data)
    else:
        print(f"Job failed: {result.error}")

Or manage the polling manually:

async with DecartClient(api_key=os.getenv("DECART_API_KEY")) as client:
    # Submit the job
    job = await client.queue.submit({
        "model": models.video("lucy-clip"),
        "prompt": "Add cinematic teal-and-orange grading and gentle film grain",
        "data": open("input.mp4", "rb"),
    })
    print(f"Job ID: {job.job_id}")

    # Poll for status
    status = await client.queue.status(job.job_id)
    print(f"Status: {status.status}")

    # Get result when completed
    if status.status == "completed":
        data = await client.queue.result(job.job_id)
        with open("output.mp4", "wb") as f:
            f.write(data)

Client tokens

Create short-lived client tokens on your backend and hand the signed token to your frontend. Its claims (service_tier, allowed models and origins, expiry, ...) are signed into the JWT, so your backend can verify and read them offline instead of round-tripping to the platform. Verification needs the verify extra (pip install "decart[verify]", adds PyJWT + cryptography):

from decart import DecartClient, TokenVerifyError, verify_client_token

async with DecartClient(api_key=os.getenv("DECART_API_KEY")) as client:
    token = await client.tokens.create(expires_in=300, metadata={"service_tier": 0})

    verified = await client.tokens.verify(token.token)  # or: await verify_client_token(token.token)
    verified.service_tier  # 0
    verified.pool          # "free" for tier 0, else "paid"
    verified.user_id, verified.organization_id, verified.api_key_id, verified.expires_at

verify checks the Ed25519 signature against the platform JWKS (https://platform.decart.ai/api/auth/jwks, fetched once and cached), plus exp, iss and aud. It raises TokenVerifyError on a tampered, expired or foreign token. It is offline JWKS verification, unrelated to the gateway's online POST /v1/verify. To inspect a token without verifying it, client.tokens.decode(token) / decode_client_token(token) returns the same fields, untrusted. The SDK is async-only; from sync code use asyncio.run(verify_client_token(token)).

Realtime fast mode

Realtime sessions accept an optional speed on RealtimeConnectOptions, alongside resolution. Fast mode (speed="fast") serves the session from a higher-compute tier for lower latency and higher throughput; output quality is unchanged. It is currently available for lucy-2.5 / lucy-latest, lucy-vton-3.5, and lucy-vton-3.6 / lucy-vton-latest (lucy-vton-latest currently resolves to lucy-vton-3.6), in the US region only, and is billed at 2x the standard realtime rate for those models. Other models ignore the option (the SDK emits a warning). Omit it (the default) for standard mode.

from decart import DecartClient, models
from decart.realtime import RealtimeClient, RealtimeConnectOptions

client = DecartClient(api_key=os.getenv("DECART_API_KEY"))
realtime = await RealtimeClient.connect(
    base_url=client.realtime_base_url,
    api_key=client.api_key,
    local_track=local_track,
    options=RealtimeConnectOptions(
        model=models.realtime("lucy-2.5"),
        on_remote_stream=on_remote_stream,
        speed="fast",  # omit for standard mode
    ),
)

Each model definition lists the speed tiers it advertises via ModelDefinition.supported_speeds (for example models.realtime("lucy-2.5").supported_speeds == ("fast",)). See the realtime docs for the full realtime API.

Development

Setup with UV

# Clone the repository
git clone https://github.com/decartai/decart-python
cd decart-python

# Install UV
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install all dependencies (including dev dependencies)
uv sync --all-extras

# Run tests
uv run pytest

# Run linting
uv run ruff check decart/ tests/ examples/

# Format code
uv run black decart/ tests/ examples/

# Type check
uv run mypy decart/

Common Commands

# Install dependencies
uv sync --all-extras

# Run tests with coverage
uv run pytest --cov=decart --cov-report=html

# Run examples
uv run python examples/process_video.py
uv run python examples/realtime_synthetic.py

# Update dependencies
uv lock --upgrade

Test UI

The SDK includes an interactive test UI built with Gradio for quickly testing all SDK features without writing code.

# Install Gradio
pip install gradio

# Run the test UI
python test_ui.py

Then open http://localhost:7860 in your browser.

The UI provides tabs for:

• Image Editing - Image-to-image edits
• Video Editing - Video-to-video edits
• Video Restyle - Restyle videos using text prompts or reference images
• Tokens - Create short-lived client tokens

Enter your API key at the top of the interface to start testing.

Publishing a New Version

The package is automatically published to PyPI when you create a GitHub release.

Automated Release

Use the release script to automate the entire process:

python release.py

The script will:

1. Display the current version
2. Prompt for the new version
3. Update pyproject.toml
4. Commit and push changes
5. Create a GitHub release with release notes

The GitHub Actions workflow will automatically build, test, and publish to PyPI.

License

MIT